Research on the Development Status of Autonomous Driving Based on End to End Technology

In recent years, end-to-end autonomous driving technology has experienced rapid advancement. Unlike traditional rule-based automated driving systems, it employs neural networks to directly map sensor data to vehicle control commands, demonstrating superior timeliness and robustness. Consequently, it is regarded as a more sophisticated solution for autonomous driving. This article first reviews the advantages and research methodologies of end-to-end autonomous driving, primarily focusing on methods based on imitation learning and reinforcement learning, while analyzing the strengths and weaknesses of each approach. Subsequently, it categorizes and discusses the testing research in end-to-end autonomous driving, predominantly centered on closed-loop testing, encompassing both real vehicle testing and open-loop testing, and compares their differences in terms of testing costs, accuracy, and safety. Ultimately, the article summarizes the current research status of major enterprises in the field of end-to-end autonomous driving. It underscores the significance of end-to-end autonomous driving technology within the realm of automated driving and provides valuable insights for future research in this domain.